Bug #21191
closed
Compute node's local keepstore doesn't seem to get the Volumes.*.PrefixLength config value
Added by Lucas Di Pentima about 1 year ago.
Updated 12 months ago.
Description
While deploying a cluster on AWS with the Volumes.<volume_name>.DriverParameters.PrefixLength config set to 3, it seems that compute node's local keepstore behave like this config is zero. I've seen blocks written with the 3-char prefix and some others without any prefix when looking at the S3 web dashboard.
- Description updated (diff)
- Assigned To set to Lucas Di Pentima
- Status changed from New to In Progress
- Status changed from In Progress to Rejected
Ok, so I've reviewed my deploy & git history on the cluster in question and my current theory is that what happened was that a-d-c didn't got restarted when the PrefixLength config change was introduced. At first I thought I added that config from the beginning but that wasn't the case.
I've restablished the PrefixLength config and manually restarted a-d-c/controller/rails services and successfully ran a diagnostic container, also checked its output log's manifest and double checked that those blocks are correctly prefixed with the correct creation date.
I'm closing this as a non-issue.
Also available in: Atom
PDF
Updated by 
Updated by